These notes are to be understood as clarifying guidelines for areas of the language that have undergone change and whose documentation is lagging behind, which has unfortunately led to a lot of speculation on how certain parts of the language currently work. This energy could be better spent on things that aren’t decided or on creative tasks.
The name of the language
The name of the language is Toaq. The tentative redefinition of the word toaq does not change this. Toaqzu is an acceptable endonym, but the language is still properly referred to as “Toaq” in English discourse. There is also a chance that the root toaq will be defined as “to be the Toaq language”. In any case, the purpose of the original re-definition of toaq was to no longer need mı when referring to the language in Toaq. Both of the possible re-definitions allow this (as either Zudeq jí Tóaqzu or Zudeq jí Tóaq).
There won’t be any “ıq”-like mechanisms
Hopefully.
There will be 1 adverbial tone (or 2, depending on the outcome of Full Arguments), which will form low-attaching adverbs when used on a member of the Main Class of predicates.
Instead of ıq/oq to create a high adverb, you can simply use the adverbial predicate as a main verb:
Lẽ bı chufaq nuo hó.
“Probably, they are currently asleep.”
becomes
Le chûfaq nuo hó.
“It is probable that they are currently asleep.”
Additionally:
There are different predicate classes. It was already mentioned that Toaq has Tense Phrases. Tense Predicate is but one of several such predicate classes. Each class can be given its own adverbial behavior. This allows words of certain classes to not be useless or “never make sense” with an adverbial tone, such as the often-cited bũ:
Nuo jí bũ.
“I am not asleep.”
Basically, many things that have no use with normal “low adverb” behavior can be given a use. That way, less is lost and ıq-like solutions can be avoided.
But also, use 
.
Phonology
While the phonology is anything but settled, certain points are somewhat clear and are worth being written down. The current situation is causing a lot of confusion, nobody knows which phonemes and word forms are canon. Some people even complain about or criticize some of the already outdated phonology changes as if they constitute the status quo, when they were really just a bunch of proposals in a sea of possibilities. You can’t know if you like something if you don’t try it, and that’s why those phonologies were tried out.
Here are some very likely properties of the final product, some of which are implicit in the latest Hoelai update:
- No obstruents in coda position.
- The schwa phoneme (written ⟨y⟩) is limited to appear in the following environments: Cy and CVy.
- The ending –uoi is no longer deprecated.
- ⟨z⟩ will stay. ⟨nh⟩ probably will, too.
- The whole idea of long vowels (in simple open syllables (CV)) is technically still undecided (though I estimate the likelihood of long vowels to be less than 50%), and depends on the below points:
The phonology can’t realistically be finalized until the outcome of the Tone Simplification proposal (known as New Segmentation) has been decided. As I said before:
“I feel that having a smaller monosyllabic root space doesn’t hurt as much when the old 
isn’t used anymore, as it’s easier to say words quickly with the new tone system”
The motivation behind all the phonology changes was to increase the monosyllabic root space. With the new tone system (outlined here http://toaq.org/Tones.pdf), I don’t mind having certain less “core” concepts pushed out to bi-syllabic roots. If we can get above 1000 possible root forms (although slightly more would be good), we’re probably close enough to functional. Long vowels (see above) could extend this space a bit further, but whether this will end up being necessary is very unclear at the moment.
New Segmentation
I’m still in favor of this proposal, but its success depends on the sandhi of the neutral tone. I haven’t tried it out enough to know how I would like it in the long run. It would be good if other people tried it out for a while to see if the neutral tone sandhi is acceptable or if it’s perceived as a nuisance (in speech).
Serial predicates, frames, and predicate families
The “old” set of serial predicate frames was nothing less than a big mess. While Hoaqgio’s serial science and frame classification brought order into an otherwise even bigger chaos, the system is still way too complicated for my taste. It is too permissive and some combinations of frames cause the cereal equivalent of spaghetti code.
The core responsibility and contribution of the serial predicate system lies in the (c 1) and (c 0) frames.
Let’s call serials involving (c 1) and (c 0) “subordinating serials”.
Subordinating serials are the most important type, but there are a few more types worth preserving:
- Adjectival serials. These are (unary) predicates expressing states and properties. They are “(c)” frames, but we should call them “(a)”, since they can invoke special adjectival semantics (let the semantics of the predicate decide if those semantics are just ru or something more nuanced (see https://toaqlanguage.wordpress.com/2019/04/08/on-the-meaning-of-adjectives/)).
- Exhibitor slots. They still feel a bit weird to me (and I was the one who proposed them), but I can’t deny that they are useful. They would be marked “e”, which would yield the frame “(c e)”.
Then there are a few miscellaneous frames, most notably SOQ, HUAQ and CHEO (these are also subordinating, but they have more places, and I would rather limit subordination to the last place, so SOQ may need to become (c c 1) or re-worked in some other way). There may still be some remaining edge cases, but in broad strokes, the list of frames should be reduced and simplified considerably as follows:
Before:
POQ: C
PAI: C C
FA: C C C
KOE: C C C 1
MIUJEQ: C C 0
HUAQ: C C 1
KO: C C 1 1
JEQ: C C 2
DUA: C 0
LEO: C 1
SOQ: C 1 C
CUA: C 1 1
MIA: C 2
NAI: 0
JIE: 0 C
GOAICA: 0 C C
CA: 0 0
PE: 0 1
JEO: 0; C 1
FOAQFUI: 1
FUI: 1 C
HAO: variable
After:
Subordinating:
TUA: c 0
LEO: c 1
???: c 2
???: c c 0
???: c c 1
Adjectival:
GI: a
Exhibitor:
??: c e
Everything else simply doesn’t participate in serial predicates (or has to be re-defined to fit within these paradigms). says Hello.
Note that predicates of Tense, Aspect, Modality and Negation technically don’t do serialization, even though it looks like they do on the surface. They simply stack from left to right. Therefore, they don’t need to be accounted for by serial frame science.
Lastly, serial predicates are right-grouping. I already know how this works in the grammar and it’s beautiful and simple and fits nicely among the other bits of this head-initial language. Left-grouping serials just don’t make sense to me (ko mi fraxu doi xy.).
Also, while the above only applies to serial frames, similar simplifications should be made to place structure frames to reduce the number of unique patterns in the language:
- Cut off all places higher than 3.
- Remove places that can be expressed using adverbs (if they are non-core places).
- Remove the FUI (1 c) place structure. Properties should come after their proprietors.
It might be a good idea to mark places that take indirect questions as well as properties with embedded interrogatives. Places that currently take a lambda to express an indirect question are changed to take an interrogative instead (which turns some “2” places into “1” places, and some “1” places into “0” places).
Numbers
With serial predicates cleaned up and due to the existence of predicate classes, it would be possible to give number predicates some kind of special behavior to make numeric expressions easier without inventing a full-fledged DSL, which I wouldn’t want (just think about Lojban’s mekso). I have no details in mind, so feel free to get creative.
If there are other topics you would like me to address, leave a comment.
The Toaq Blog 
> Note that predicates of Tense, Aspect, Modality and Negation technically don’t do serialization, even though it looks like they do on the surface. They simply stack from left to right. Therefore, they don’t need to be accounted for by serial frame science.
Can you expand a bit on this? Mainly, does this mean that when predicates of these classes are used in “serials”, they should be thought of as simply their corresponding operators ¬, □, etc. rather than their predicate forms?
LikeLike
1. I don’t consider there to be a difference between an operator and a “normal” predicate. To quote And Rosta: “Presupposing that everything effable (i.e. linguistically expressible) can be represented as a predicate–argument structure, a logical language (in the technical sense, i.e. loglang) is one that unambiguously bidirectionally encodes an unlimited number of PASs. (What Pycyn terms ‘monoparsing’. The bidirectionality is going from PAS to phonological form and from phonological form to PAS.)”
If you fully abstract everything else away, you are only left with predicates and arguments. From this perspective, negation is just as much a predicate as a modal operator or an illocution operator is. Whether you represent the abstract [Neg] as “bu” or ¬ in your meta-language should make no difference with regards to the truth conditions of this predicate. So think of it as ¬, but not just “when used in serials”.
2. I cannot lay out the entire syntax here. On the surface, there is no difference at all between Neg in front a predicate (or serial predicate) and a Neg that directly takes a noun phrase; it behaves identically to an ordinary predicate (specifically one with frame (0)). Having certain predicates in specialized classes, even if their distribution is seemingly the same as that of a normal predicate, helps keep things tidy, give them their appropriate semantics (some classes may require different hidden variables than others for example), and make it possible to give them special adverbial grammar.
LikeLike
Re 1: I understand that ¬ and predicate-bu probably look very similar for practical purposes and also from the perspective of the grammar, however when doing formal logic it is usually very important to distinguish between them. This is because while ¬ can only take formulas, predicate-bu can take anything from propositional constants to quantified variables, and is for that reason far stronger. (Strong enough that attempting to equate the two typically results in contradiction!)
That’s why I’m concerned with which of these two things bủ actually represents. Certainly it has to be interpreted as predicate-bu at least some of the time, because it can gramatically take quantified variables, but the question is whether in the case of serials and the like it should be understood as automatically reduced to ¬ “by the grammar”, or if it’s just up to our logical inference skills to make that simplification in our heads.
LikeLike
> Each class can be given its own adverbial behavior. This allows words of certain classes to not be useless or “never make sense” with an adverbial tone
If words of different classes are mixed together in a serial and used as an adverbial, how would the adverbial’s behavior be determined?
LikeLike
The basis rule is that the the highest predicate determines the type of the whole phrase and therefore its adverbial behavior. With some predicates (call them weak predicates), it is more useful to say that the highest *strong* predicate determines the type. All non-Main Class predicates are strong. Some Main Class predicates are weak, such as, e.g. “jaq”. For example, if “lẽ” (“le” being a Modal) attaches high, then so would “jãq le”. It would be less useful if it had to attach low.
LikeLike
Do you have any comment on the relationship between t5 clauses, Davidsonian events, and the “expansion” of low adverbs (if that even makes sense)? While it was not addressed in the post I thought I’d ask, as this would be the biggest question keeping me from using modern Toaq.
For example in the sentence “Jỉa nûo jí”, what does the t5 clause stand for? Is it a proposition, an event, or something else?
LikeLike
Low adverbs are meant to be understood in a (neo-)Davidsonian way: tảo jí hão -> ∃e. tao(e) ^ AGENT(e, ji) ^ hao(e). The event is not an intrinsic part of the argument structure (which is why adverbs don’t introduce their own event variable). Instead, it is introduced by a higher, separate head, and depending on the analysis, it can be helpful to say that certain types of verbs don’t combine with event arguments (the stage-level vs individual-level distinction is related).
As for the second question, which is a very loaded question, t5 clauses are CPs (Complementizer Phrases), so the question is what CPs denote. A (the?) standard theory of CPs (at least of that-clauses) is that they are propositions, and this is more or less how I thought about it when designing the language. Note that t5 clauses have no syntax for quantification, which makes sense for (singular) propositions, but would be unfortunate if t5 clauses denoted events, which are suitable to quantification (you can do that via “guo” or “tue”). However, another interesting theory is that CPs are predicates of propositional content, i.e., instead of denoting sets of possible worlds (one of the standard views on propositions), they denote properties (sets) of individuals carrying propositional content. See Kratzer 2006, Moulton 2009, 2015 for more details. I can’t say that I have made up my mind, but I don’t think t5 clauses denote events, at least.
LikeLike
Okay, then it looks like my understanding of low adverbs alone was correct. (Yay!)
As for content clauses, an issue I see with them not being events (not even baq event) is that it means one of three things has to happen:
1. You can’t use rao with events.
2. You can’t use rao with t5 clauses.
3. rao is polysemous.
None of these seem like attractive options. This issue was the inspiration for my “baq event” interpretation, which I’ve been using recently and which I feel has been working pretty well (though I haven’t used it much). Under this interpretation, you can get cool stuff like the following (using “lo” to mean “guo” but in PO):
1. Cả hôaq na gîo da. “Burnings cause there to be light.”
2. Cả tu lỏ hôaq na sa lỏ gîo da. “Every burning causes there to be some light.”
3. Bủ câ ní lỏ hôaq na gîo da. “This burning didn’t cause there to be light.”
With t5 clauses not being baq event, it’s hard for me to see how you wouldn’t need Q lỏ in front of every single one of them. (Which is why I proposed alternatively that t5 clauses could be their own thing but that they could result in constructs that have the same grammar as predicates in t4, so that you could just put quantifiers directly in front of them without needing a lo.)
LikeLike
The original design was to give t5 clauses as little pre-baked semantics as possible, because the goal was that the same t5 clause should be able to co-fill lots of different places (pure propositions, events, …), and the idea was that it is the predicates which take t5 clauses that decide how a t5 clause is to be interpreted in that specific argument place.
Any formalization that leads to you not being able to say “rào tî jí máq” goes against that philosophy.
LikeLike
(I’m not sure if I’m replying to the right comment here, I’m so not used to threads that can’t nest infinitely :o)
I agree about “rào tî jí máq”, of course. But to me, this seems like an argument in favor of t5 clauses being baq event! Becuase surely you also want to be able to say “rào ló tî jí máq” or “rào sáqkō lỏ tî jí máq”, which clearly reference individual events. So if t5 clauses are their own thing, then either (a) one of these two uses will have to be forbidden, or (b) rao and many other predicates will be polysemous.
On the other hand, the “baq event” interpretation seems to handle all of these cases well. I haven’t yet thought of a proposition-taking predicate that doesn’t work just as well if its “proposition” is framed as an event-Kind instead. For example, under this interpretation, **bu** is a Kind predicate which means something like **lı gủo sıa ja**. The word **chi** would also be a Kind predicate (on its second slot).
**rao** on the other hand makes sense to apply to individual events (instead of just Kinds) and would be individual-level or stage-level. Individual-level seems to make the most sense as two events don’t really start or stop being simultaneous, but the paraphrasability with “sa” seems desirable. I’d have to think about this some more.
And on the other side of the scale, for the interpretation where t5 aren’t events, not even baq event, we still have the polysemy thing I pointed out above, which you didn’t address just now unless I’m misunderstanding you.
LikeLike
And to put the case even more succinctly: the mere fact that the rao in “rào tî jí máq” is in the adverbial tone means that its FIRST argument is an actual event!!
LikeLike
Okay, Seoqrea has shared some extremely interesting thoughts inspired by your use of the term “complementizer phrase”. We didn’t manage to get their idea copied into my head at a very high fidelity, but it sounds like the gist is that Toaq gets to “know” where a t5 clause is before deciding what to do with it. So if it’s the second argument of dua, then it will simply take the proposition directly expressed by the clause and pass that to the predicate, whereas if it’s the first argument of rao, then it first instantiates the proposition into an event (with ke or sa or baq or some such) and then passes that event to rao, and rao never has to “see” anything other than the events it’s used to. Is that right or on the right track?
LikeLike
> I probably forgot to address some important things. If so, leave a comment.
Here’s one thing I’d be curious about. I’ve had this headcanon about Toaq for a while that t5 means “baq event”, where “events” are the things that are quantified over in the formula:
∃e. suaq(e, hó) ^ de(e)
which is (what I understand to be) the transcription into logical form of the Toaq sentence Sủaq hó dẽ da. Individual events could then be recovered with “guo” as in “ke güo jâı”.
My question is: does this headcanon make sense to you? If it makes sense, does it “resonate” with you and/or does it match the way you think about things?
If it doesn’t make any sense, but explaining why would be difficult or burdensome, feel free to just say “This doesn’t make sense to me” or something — I promise not to interpret the terseness as rudeness ^^
LikeLike
See my response to Robin.
LikeLike
Here are my thoughts, sorted by section:
— The name of the language —
Cool. It seems most intuitive to me that inside the language, “toaq” should simply be a culture predicate “___ pertains to Toaq culture, etc” and that “toaqzu” would be a predicate “___ is the toaq langugae”, following the pattern of “hiqli(zu)” and all other cultures. I will likely use the toaqzu endonym a lot as it’s a ripe candidate for a random Toaq word to sprinkle into English text. Er, I mean a ripe Toaqzu word.
— There won’t be any “iq”-like mechanisms —
Noooooo, my favorite toy!
Oh well, it does seem potentially more streamlined that the main claim of a sentence should always be its verb rather than sometimes being an iq-construct, and that correspondingly subordinate stuff would always be done with t5.
This predicate class stuff is interesting. You mention in a different section that tense, aspect, modality, and negation will no longer technically be serials but rather simply stack left to right, and will be their own predicate class(es?). I’ll be curious to see exactly what this cashes out to. I’m imagining something like:
Bủ dủa jí hóq lũı
Will work out to be:
bu(lui( dûa jí hóq ))
Or maybe the same thing with bu and pu swapped.
— Phonology —
All rise for the return of the honorable -uoi ending! No other syllable could ever have fit the current definition of kuoi as well as kuoi does.
I’m very curious to see whether the “current” phonology will be enough, supposing no plosive codas (as you state is likely) and no long vowels (as you speculate). I’ve been trying to coin a bunch of missing common words by simply looking around me and seeing what things I can’t describe in Toaq, for this exact reason: I want to know how far I can get before I discover that there aren’t enough short words to go around, with one possible answer being “you can go as far as you like and that won’t happen ever”. I’ll keep people posted on this — you’ll know how it’s going by the ratio of times when I coin a common word to times when I complain about not being able to think of an etymology for a common word.
I’ll miss -yq, especially since I imprinted on my zi/dyq pair almost as soon as I coined it, but c’est la vie.
I agree with your evaluation that New Segmentation would make long words easier to say, and that pushing some stuff out to bisyllabic space becomes more acceptable. Like “mutē” for “trash”, which I also just like the sound of.
— New Segmentation —
As I said in #general, I’ve been using the New Segmentation in everything, and I feel like I’ve gotten used to the neutral tone sandhi fairly fast. I’ve mostly been rendering illocutions as old t6 when possible and current t7 when not. I think it sounds fine, and so far has not felt like a nuisance.
How did you like the sound of it in my recording of the Go article?
— Serial predicates, frames, and predicate families —
“Hoaqgio’s serial science” ❤️
I agree with your sentiments about spaghetti code serials. Limiting subordination to the last place is a very attractive option to me. I like to think about serializing workhorses (like leo and huaq) in terms of the action they have on the front of a predicate’s place structre. leo, being a (c 1) predicate, leaves the number of slots alone but changes the relationship of the first slot to the verb and remaining slots. huaq, being a (c c 1) predicate, adds one slot, essentially doubling the first slot so you can make two sentences that share a tail and then compare them. Predicates that subordinate elsewhere are harder to think about in this fashion (though of course you can do it).
I was surprised to find that you were the one who proposed exhibitor slots, as I’d had it filed in my mind as a uakci proposal.
I notice you put ??? for the names of the frames (c 2), (c c 0) and (c c 1). These frames did have names in the old convention. At first I wondered if this was a sign you didn’t like the community definition of mia (“___ is in relation ___ with many things”), but I have a hard time imagining that the same thing applies to huaq and miujeq. Is this a typo?
I suggest MUOQ as the name of the (c e) frame.
Where is jie going? Surely not to mù mỉu? So then probably to miu + t5 (?).
I admit I’m relieved at the confirmation that serials will remain right-grouping. I was apprehensive about having to relearn my whole toolbox of tricks for thinking about them. It would have been an interesting task, though. Good thing there’s no shortage of those in this project!
> Places that currently take a lambda to express an indirect question are changed to take an interrogative instead
Woo! This is my aesthetic druthers.
> (which turns some “2” places into “1” places, and some “1” places into “0” places).
Hmm. Words like **mieq** when used as serial heads still need to consume a slot from the tail. Can they do that if they’re (c 0) frame? Maybe some kind of “0i” place can be invented, and we can ~~make serials complicated again by~~ having a MIEQ frame as (c 0i).
LikeLike
Thank you for the comments.
I still have to listen to the Go article.
I simply didn’t feel like spending time on deciding which predicate most deserved to be the eponymous member of its class. I agree about MUOQ.
Are you asking about “JIE”‘s fate in serials or about JIE-like place structures in general? JIE is the way it is because its main use was meant to be as an adverbial. However, it doesn’t work as a normal low-attaching adverb. I don’t know if I want to have a JIE predicate class for related sentence adverbs (aka high adverbs).
LikeLike
Sounds good, Hoemai.
What I was asking about with jie was mainly the fate of the jìe X construct, which I agree doesn’t work as a low adverb (“She thinks it’s beautiful” != “It’s beautiful and she thinks the event of that”). A predicate class for JIE (and related predicates?) could be neat, I’m tentatively in favor. Actually, maybe jie itself would be the only member of that class, with other semantics accessible via serials: jìe lẻ jí, jìe dẻ hó.
LikeLike
⸨For some unfathomable reason, the ʀᴇᴘʟʏ link doesn’t appear under the message I wanted to react to, therefore I’m doing so here.⸩
With regard to the ⟪rao⟫ argument type issue, if all you’ve got is an event and not a proposition, and you want to apply ⟪rao⟫ on it, you may convert it back to a proposition using ⟪guo⟫. But a more succint method would presumably being having a sibling predicate to ⟪rao⟫, taking as argument spatiotemporal entities such as events instead of abstract propositions. ⟪raoguo⟫ or ⟪guorao⟫ could be possible candidate forms.
As for the idea of having ◌̂ clauses behaving like ◌̉ words, that doesn’t sound like a good idea to me, as you’d always need to put a quantifier in front of it —you can’t stack the raising tone ◌́ onto ◌̂ after all!—, making things significantly more verbose.
LikeLike
I don’t see how you can use guo to turn an event into a proposition, since an event can guo multiple different propositions. But you might be able to use faq, as even if hóq refers to an event, fâq hóq still refers to whatever a t5 clause refers to, which could be a proposition.
I find the duplicated predicates an unaesthetic solution, as you’d need to duplicate every single predicate that could take either a proposition or an event. In fact, a proposition with N slots that can take either one would need to be duplicated 2^N times.
As for the quantifiers idea, I agree about the verbosity.
LikeLike
┌───────────
│ I don’t see how you can use guo to turn an event into a proposition,
│ since an event can guo multiple different propositions.
└───
Isn’t the issue that propositions may mu guo several different events, rather? In which case, ⟪tủq mủ gủo⟫ might come in handy.
LikeLike
Well, I was imagining you meant that an event could be turned into a proposition via a phrase like “ke güo [event] hóa”, and I was thinking this would only work if guo mapped every event to just a single corresponding proposition. Otherwise, you’d have issues with (1) getting back more propositions than you want and (2) those propositions also describing other events.
So yeah, I guess it’s also an issue that a prop can mu guo more than one event.
But in any case, I now think that this conversion thing isn’t an issue, as Seoqrea and Liq have suggested that t5 could be “versatile” in such a way as to produce events when events are needed and props when props are needed, without any predicate being polysemous.
LikeLike
(And reading back Hoemai’s first reply to my complaint of polysemy, it sounds like what these two describe is what Hoemai had in mind, even though I didn’t understand the idea when I first read Hoemai’s reply.)
LikeLike